It is by nowwell established that children as youngas twoyears of ageare sensitive to the knowledgeof their interlocutorsin selecting appropriate linguistic forms to realize subjects and objects in spontaneous speech (see review in Allen et al.,2008). In particular, theyare sensitive to the conceptual accessibility of a referent, definedas ‘‘the easewithwhich thementalrepresentation of some potential referent can be activated in or retrieved from memory’’ (Bock andWarren, 1985:50). Thisconceptual accessibility is determined by a set of discourse-pragmatic features including PRIOR MENTION (whether the referentis new to the discourse or has already been introduced), PHYSICAL PRESENCE (whether the referent is absent from or present inthe physical context), and JOINT ATTENTION (whether the speaker and interlocutor are jointly attending to the referent at the timeof speech or not) (e.g., Chafe, 1987; Du Bois, 1987; Givon, 1983; Gundel et al., 1993; Prince, 1985). Speakers then direct
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
* Corresponding author. Tel.: +1 15087889267.** Corresponding author at: Psycholinguistics and Language Development Group, Faculty of Social Sciences, University of Kaiserslautern,Erwin-Schrödinger-Straße 57/409, 67663 Kaiserslautern, Germany.
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx2
+ ModelsLINGUA-2160; No. of Pages 19
their interlocutors’ activation or retrieval of the intended referents by using particular forms in speech to signal the level ofaccessibility of the referents (Ariel, 1994:99). Referents that are not accessible to the interlocutor -- for example, referentsnewly introduced into the discourse -- are typically realized with maximally informative forms like lexical noun phrases as inSpeaker A’s utterance in (1). In contrast, arguments that are already accessible to the interlocutor through one or morediscourse-pragmatic features -- for example, referents mentioned in the preceding utterance -- are typically realized withreduced forms like pronouns, as in Speaker B’s utterance in (1).
(1)
Pleon
1 Aages
Speaker A:
ase cite this areferential ch
ll studies are witof 5;09--15;08.
Good morning, Bob. What a nice shirt you’re wearing!
Speaker B: It doesn’t fit me anymore.
This process of referential choice reflects sophisticated coordination of linguistic knowledge (which linguistic forms arepermitted in a given language), discourse knowledge (how accessible referents are in the discourse and physical context),and social knowledge (what the interlocutor is attending to at the time of speaking).
In investigating this process, previous researchers have examined discourse-pragmatic features individually,compared the interaction of two or more discrete features, or looked at the effect of accessibility in general withoutdifferentiating individual features (e.g., Allen, 2000; Clancy, 1993, 1997; Guerriero et al., 2006; Hughes and Allen, 2013;Narasimhan et al., 2005; Serratrice, 2005; Skarabela et al., 2013; see Allen et al., 2008 for a complete review). In thesestudies, researchers code spontaneous speech or experimentally test one or more individual discourse-pragmaticfeatures. They then assess children’s sensitivity to each feature in isolation and, if more than one feature is tested, oftendetermine which one of the various features children are most sensitive to. However, it is very unlikely that children in factattend to each individual discourse-pragmatic feature in isolation. Rather, they almost certainly attend to the variousfeatures in interaction and in combination with each other.
This sensitivity to the interaction between features has been much less well studied than sensitivity to the features inisolation even though the former is a better approximation of how things happen in the real world. Only one study thus far(Allen, 2007) has examined the incremental effect of accessibility by adding features together, one at a time in a stepwisemanner, in order to investigate whether the resulting increase in ‘‘inaccessibility’’ determines the use of a high informationforms, such as a lexical NP. Allen (2007) found positive results using this method in a null subject language, Inuktitut, butshe used only a few features in a relatively small data set. The goal of the present study is to test this method morethoroughly, by including more accessibility features and using data from a language that does not permit null subjects. Inthis way, the present paper seeks to extend the literature on the children’s sensitivity to the interaction between features.
1.1. Previous approaches to studying the interaction of features
Numerous researchers who have investigated children’s acquisition of referential choice have found that preschoolchildren are sensitive to a number of individual features of discourse and social knowledge. These include the threefeaturesmentioned earlier (PRIOR MENTION, PHYSICAL PRESENCE, JOINT ATTENTION) as well as several others including CONTEXTUAL
DISAMBIGUATION (whether a particular referent has potential competitor referents in the physical context), LINGUISTIC
DISAMBIGUATION (whether a particular referent has potential competitor referents in the linguistic context), ANIMACY (whetherthe referent is animate), TOPICALITY (whether the referent is the focus of conversation), QUERY (whether the referent is thesubject of or the response to a question), EXPLICIT CONTRAST (whether a contrast is made explicit or emphasized by thespeaker), and PERSON (whether the referent is first/second vs. third person) (see Allen et al., 2008). Further, this sensitivityto referent accessibility is not restricted to one language or one language typology. Similar results have been found innumerous languages1 including English (Campbell et al., 2000; Graf, 2010; Guerriero et al., 2006; Gundel et al., 2007;Hughes, 2011; Hughes and Allen, 2013; Matthews et al., 2006; Mishina-Mori, 2007; Serratrice et al., 2004), French (DeCat, 2004; Salazar Orvig et al., 2010a, 2010b), German (Schmitz, 2007; Wittek and Tomasello, 2005), Hindi (Narasimhanet al., 2005), Inuktitut (Allen, 2000; Allen and Schröder, 2003; Skarabela, 2007a, 2007b; Skarabela et al., 2013), Italian(Serratrice, 2005; Serratrice et al., 2004), Japanese (Guerriero et al., 2006; Mishina-Mori, 2007), Korean (Clancy, 1993,1997), Mandarin (Huang, 2011; So et al., 2010), Spanish (Paradis and Navarro, 2003; Shin and Cairns, 2012) and Turkish(Demir et al., 2012; Gürcanli et al., 2007). Overall, preschool children are sensitive to a variety of individual discourse-pragmatic features in choosing whether to use an informative or reduced form to realize a referent in speech, and they areable to integrate that sensitivity with the language-specific possibilities for referential choice.
As research in the field progressed, it became increasingly clear that children were not only sensitive to these featuresindividually, but they were sensitive to the interaction of these features. Several different approaches have been used so
rticle in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityoice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
h children aged 2--4 years except for Shin and Cairns (2012) who elicited data from Spanish-speaking children between the
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 3
+ ModelsLINGUA-2160; No. of Pages 19
far to gain insight into the sensitivity to this interaction between features. One approach has been to assume from theoutset that the features are attended to in combination and thus to simply treat them as one feature. Serratrice (2005)follows this approach in her study of Italian-speaking children between the ages of 1;7 and 3;3. She assesses the featureACTIVATION which measures the degree of ‘‘identifiability and accessibility’’ of a referent (p. 440) -- essentially a combinationof PRIOR MENTION, JOINT ATTENTION to the referent, and TOPICALITY. Clark and colleagues discuss the influence of commonground on referential choice for French-speaking children at 2;3 and 3;6 (Clark and Bernicot, 2008; Clark and Amaral,2010). Common ground refers to the information that is ‘‘mutually known’’ between speaker and interlocutor (E. Clark,2009:282; see also H. Clark, 1996), and thus is also essentially a combination of PRIOR MENTION, JOINT ATTENTION to thereferent, and TOPICALITY. This approach has the shortcoming of not being very fine-grained -- it is clear that the referents areaccessible, but it is not clear what in particular makes them accessible.
A second approach has gone in the opposite direction, statistically assessing the independent contribution of each ofseveral features (Allen, 2000; Hughes, 2011; Hughes and Allen, 2013; Skarabela, 2007a). For example, Allen (2000), whostudied Inuktitut-speaking children from ages 2;0 to 3;6, found using logistic regression that an explicitly contrastedreferent had a greater likelihood of being realized with an informative form (i.e., demonstrative or noun phrase) than areferent newly introduced into discourse, which in turn had a greater likelihood of being realized with an informative formthan a referent with potential competitors in the physical context. This approach is somewhat problematic, however, giventhe difficulty of having enough data points in spontaneous speech data to meet the requirements of logistic regression.
A third approach has been tomathematically assess the combinatory effect of the different discourse-pragmatic features.In this approach, the important factor is the number of discourse-pragmatic features that a particular referent is accessible ornot accessible for. It is implicitly assumed that eachof the featureshasanequivalent effect on referential choice, andso it is ofless importanceexactlywhich featuresa referent isor isnot accessible for.Asanexample, take the referent ‘‘shirt’’ in (1). Thisreferent is introduced in Speaker A’s utterance where it is ‘‘accessible’’ for four of the previously mentioned discourse-pragmatic features (PHYSICAL PRESENCE, LINGUISTIC DISAMBIGUATION, QUERY, EXPLICIT CONTRAST) and is ‘‘not accessible’’ for six ofthem (PRIOR MENTION, JOINT ATTENTION, CONTEXTUAL DISAMBIGUATION, TOPICALITY, ANIMACY and PERSON). Therefore, its inaccessibilityscore is 6 out of 10 -- a relatively high score that leads to its realization as a noun phrase by Speaker A. For Speaker B’sutterance, this same referent is now ‘‘accessible’’ for eight of the previously mentioned discourse-pragmatic features (PRIOR
CONTRAST) and is ‘‘not accessible’’ for two of them (ANIMACY and PERSON). Therefore, its inaccessibility score is 2 out of 10.Because this is a low score, the referent is realized as a pronoun in Speaker B’s utterance.
The simplest mathematical method is to assume a ‘‘threshold’’ of accessibility -- a point below which children assumethat a referent is more accessible, and above which they assume that it is less accessible (Allen, 2000; Narasimhan et al.,2005). In their study of Hindi-speaking children’s sensitivity to four discourse-pragmatic features (ages 3;0--4;0),Narasimhan et al. (2005) found that children differed in their referential choice for referents that were pragmaticallyprominent (i.e., ‘‘not accessible’’ for any one or more of the four features) vs. those that were not pragmatically prominent(i.e., ‘‘accessible’’ for all four features). Thus, their threshold was set at the point of ‘‘not accessible for one feature’’ (i.e., 0vs. 1+ features). Allen (2000) tested two thresholds: one set at ‘‘not accessible for one feature’’ (i.e., 0 vs. 1+ features), andone set at not accessible for two features (i.e., <2 vs. 2+ features). Children’s referential choice showed sensitivity to bothof these thresholds.
A second mathematical method is to explore children’s sensitivity to the interaction between two particular featureswhile controlling for the effect of other features. For example, Skarabela and colleagues (Skarabela and Allen, 2002;Skarabela et al., 2013) used Inuktitut spontaneous speech data from children ages 2;0--3;6 and coded for eight discourse-pragmatic features (PRIOR MENTION, PHYSICAL PRESENCE, JOINT ATTENTION, CONTEXTUAL DISAMBIGUATION, LINGUISTIC DISAMBIGUATION,ANIMACY, QUERY, EXPLICIT CONTRAST) to explore the interaction between two of the features: PRIOR MENTION and JOINT ATTENTION.To hold the effect of the other six features constant, only referents that were coded as ‘‘accessible’’ for all of those featureswere included in this analysis. When the effect on referential choice of each of PRIOR MENTION and JOINT ATTENTION wasassessed separately, the two features had a similar magnitude of effect: lexical noun phrases were used to realize 21% ofreferents that were ‘‘not accessible’’ for PRIOR MENTION and 22% of referents that were ‘‘not accessible’’ for JOINT ATTENTION. Incomparison, lexical noun phrases were used to realize only 8% of referents coded as ‘‘accessible’’ for PRIOR MENTION andonly 4% of referents coded as ‘‘accessible’’ for JOINT ATTENTION. However, the children were very sensitive to thecombinatory effect of the features: 64% of referents coded as ‘‘not accessible’’ for both of the two features were realizedlexically compared to only 3% of referents coded as ‘‘accessible’’ for both features.
A more sophisticated mathematical method is determining the incremental effect on referential choice of a whole set ofdiscourse-pragmatic features. As in the two other mathematical methods described, a referent that is ‘‘accessible’’ for allfeatures is least likely to be realized with an informative linguistic form (i.e. demonstrative or noun phrase). The likelihoodof being realized with an informative form increases with each additional feature for which the referent is not accessible.Thus, a referent that is coded as ‘‘not accessible’’ for four features is more likely to be realized with an informative formthan a referent that is coded as ‘‘not accessible’’ for only three features, etc. Allen (2007) tested this premise in
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx4
+ ModelsLINGUA-2160; No. of Pages 19
spontaneous speech data from Inuktitut-speaking children ages 2;0--3;6 using four features which were found to havesignificant effects on referential choice in previous studies (PRIOR MENTION, PHYSICAL PRESENCE, EXPLICIT CONTRAST, andCONTEXTUAL DISAMBIGUATION). A clear and subtle relationship between accessibility and referential choice was found asshown in (2). (Note that informative forms here include demonstratives and noun phrases, and that no referents werecoded as ‘‘not accessible’’ for all four features.)
(2)
Pleason re
a.
e cfere
referents ‘‘accessible’’ for all four features! 18% realized with informative forms
b. referents ‘‘not accessible’’ for one feature ! 29% realized with informative forms c. referents ‘‘not accessible’’ for two features ! 57% realized with informative forms d. referents ‘‘not accessible’’ for three features ! 86% realized with informative forms
These results show that children are not only sensitive to a threshold of features or to the effect of two features incombination, but are also sensitive to the relatively more subtle incremental effect of features.
This incremental approach is very promising for determining just how subtle children’s sensitivity to accessibility is.However, it has only been tested in one language and should ideally be replicated in other languages of differenttypologies. Importantly, Inuktitut allows referents to be realized as null forms (i.e., the referent is omitted), and it does notallow pronouns to appear in either subject or object position. Therefore, the difference between a reduced form (null form)and a more informative form (demonstrative or lexical noun phrase) is maximally clear. It is not yet known whetherchildren’s sensitivity would be so subtly revealed in a language like English where null forms are not grammatical and thetypical reduced form is a pronoun. Further, Allen’s (2007) study did not assess the incremental effect of accessibility incaregiver data, and did not look for any effect of development.
1.2. The present study
The purpose of the present study is to fill these gaps in the literature. We assess the incremental effect of six discourse-pragmatic features in argument realization in a set of spontaneous speech data from four English-speaking children at twotime points (2;0--2;7 and 3;0--3;1), aswell as in data from their caregivers. The role of child-directed speech in the acquisitionof referential forms has already been established in several previous studies. For example, Küntay and Ӧzyürek (2006)examined the production of demonstratives in monolingual Turkish-speaking adults and children. Their results showed apattern of demonstrativeusewhenadults spoke to children asopposed to speaking to other adults, and the childrenmirroredthis non-adult-like pattern. In another study, Guerriero et al. (2006) compared monolingual English-speaking mothers andchildren tomonolingual Japanese-speakingmothers and children. The researchers found different patterns of developmentin theproductionof referents forEnglish-acquiringandJapanese-acquiring children.Thesedifferences reflectedadifferencein input between the two groups of parents, both in choice of referents and in terms of interactional style. Rozendaal andBaker (2010) investigatedEnglish-acquiring children’s acquisition of determiners and pronouns in terms of the given vs. newdistinction in discourse, and found that the children’s development of the determiners occurred at a faster rate than theirpronoun use. This appeared to be a direct effect of the adult input data, which revealed inconsistency in pronoun usage. Inlight of these findings, we believe it is important to consider the shape of the input data in this study in order to investigatepossible parallel patterns in terms of incremental effects of accessibility in the adult and child data.
We have already found in earlier studies that the children in the present study show sensitivity in their referential choiceto each of six individual discourse-pragmatic features (Hughes and Allen, 2006, 2013). Here we build on that study byassessing the effect of these same features incrementally. We hypothesize that English-speaking children will showincremental sensitivity to discourse-pragmatic features in the same way as Inuktitut-speaking children, regardless of thedifferences in language typology. We also expect that caregivers will show high sensitivity to the incremental effect ofdiscourse-pragmatic features. Finally, we expect that the three-year-old children will show a more sophisticatedincremental sensitivity than will two-year-olds.
2. Method
2.1. Participants
Data for this study were taken from the Manchester-Max Planck Dense Database (Lieven et al., 2003, 2009).Participants were the same four monolingual English-speaking children studied in Hughes and Allen (2013) -- Annie,Brian, Eleanor, and Fraser. The children lived with their families in a large metropolitan area in England and came frommiddle-class backgrounds. All four children were audiotaped for several hours each week and videotaped for one houreach week at ages 2;0--2;1 and 3;0--3;1. One child, Brian, was also taped throughout the entire year. For this study, we
ite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityntial choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 5
+ ModelsLINGUA-2160; No. of Pages 19
used only the videotaped data (one hour per week per child) so that non-verbal aspects of the interaction and the physicalcontext could be included in the analyses.
2.2. Data
The transcripts and videotapes, which were generously made available to us by Elena Lieven, recorded the children inspontaneous interactions with their mother and/or a familiar researcher (see Lieven et al., 2003, 2009 and Hughes andAllen, 2013, for further details about the recording, transcription, and coding process). In order to assess the children’sdevelopment and based on previous findings that children’s sensitivity to accessibility changes over time (e.g., Hyams,1986; Guerriero et al., 2006; Serratrice et al., 2004; Salazar Orvig et al., 2010a,b), the children’s utteranceswere assessedat the two available time periods. The ultimate goal was to analyze the children’s utterances when they were all at similarlinguistic levels. These two time periods will be referred to throughout the remainder of this article as Time 1 (2;0--2;1 for 3children and 2;4--2;7 for 1 child2) and Time 2 (3;0 and 3;1 for all four children). From the available data, the following wereexcluded: utterances that were not fully intelligible; utterances that did not contain a verb; and utterances that were exactself-repetitions, exact imitations, recitation of poems or songs, frozen forms, and routines. This yielded a total 1901utterances from 25 files at Time 1 and 3177 utterances from 21 files at Time 2. These 5078 utterances were then coded forgrammatical and discourse-pragmatic information as described in the following sections.
In the analysis, we focused on the effect of referent accessibility on referential choice. Therefore, we restricted the dataset for analysis to contexts in which the accessibility of the referent in the discourse context could be determined to havean effect on the linguistic form produced by the speaker as opposed to the linguistic form being fixed for other reasons.This led to excluding the utterance types listed in (3):
(3)
Pleason re
2 Onewe alsochildren
3 At Tsignificaomittedrespect
a.
e citeferent
child, Bhad lawouldime 1, tntly hig16% oively), w
imperative utterances, because subjects of imperatives must be omitted for grammatical reasons in English(Time 1: 227; Time 2: 317)
b.
utterances containing first and second person subjects, because first and second person referents are bydefinition virtually always fully accessible in the discourse (Time 1: 627 first person and 60 second person;Time 2: 1194 first person and 278 second person3)
c.
utterances in which children and adults use proper names to refer to themselves or their interlocutors insteadof first and second person pronouns (Time 1: 48; Time 2: 88; see Hyams, 2008, and Hughes and Allen, 2013for more explanation)
d.
utterances in which the existence of JOINT ATTENTION could not be determined due to the positioning of thevideo camera or movement out of frame by the participants (Time 1: 67; Time 2: 118)
e.
utterances containing subjects that are explicitly contrasted with an alternative referent using stress or toneof voice (e.g., ‘This cup is mine and that cup is yours’) because this feature has been shown in previousstudies to be categorical in English, necessitating rather than simply influencing the use of an overt subject(Time 1: 25; Time 2: 57)
f.
utterances containing subjects that are the referent of or the answer to a question (e.g., referent of question:‘What broke?’; answer to question: ‘The vase broke.’) because this feature has been shown in previousstudies to be categorical in English, necessitating rather than simply influencing the use of an overt subject(Time 1: 124; Time 2: 97).
By including only utterances in which argument choice could truly be affected by a referent’s accessibility in discourse,our analysis provides a clearer picture of the actual effect of the six discourse-pragmatic features in this study.
The numbers for the final set of data used for analysis are given in Table 1, in which the age, number and length ofsessions, the MLU, and the number of possible subject contexts for each child at each of Time 1 and Time 2 are listed.
In order to examine the state and effect of the caregiver input, 15% of the utterances from the four mothers wererandomly selected and coded using the same criteria as outlined above for the child data, which yielded a total of 1129possible subject contexts for the four mothers.
this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityial choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
rian, had a somewhat lower mean length of utterance (MLU) than the other three children during these earliest recordings. Sinceter recordings for him, we used a slightly later set of data for his Time 1 -- from 2;4 to 2;7 -- so that the linguistic level of all fourbe as comparable as possible.he children omitted subjects in 37% of all first person utterances and 60% of all second person utterances. These percentages areher than the percentage of omitted third person subjects at Time 1 (24%). At Time 2, the children omitted far fewer subjects but stillf all first person subjects. This is significantly higher than the proportion of second and third person omissions (3% and 4%hich show a more adult-like rate of omission.
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx6
+ ModelsLINGUA-2160; No. of Pages 19
Table 2Accessibility features.
Accessibility feature Accessible value Inaccessible value
ANIMACY (AN) Animate InanimateCONTEXTUAL DISAMBIGUATION (CD) Only referent in physical context Multiple referents in physical contextPHYSICAL PRESENCE (PP) Physically present Physically absentPRIOR MENTION (PM) Given: mentioned within
preceding 5 utterancesNew: not mentioned withinpreceding 5 utterances
LINGUISTIC DISAMBIGUATION (LD) No other possible referents inpreceding 5 utterances
Other possible referents inpreceding 5 utterances
JOINT ATTENTION (JA) Referent focus of attention forchild and interlocutor
Referent not focus of attention forchild and interlocutor
Table 1Age, mean length of utterance in words, and number of 3rd person subject contexts produced by four children at Time 1 and Time 2.
Annie Brian Eleanor Fraser
Time 1 Age 2;0.4--2;1.22 2;4.17--2;7.1 2;0.4--2;1.8 2;0.1--2;1.10Sessions 7 � 1 h 6 � 1 h 6 � 1 h 6 � 1 hMLU 2.15--2.45 1.88--2.22 1.68--2.17 1.35--1.74N of Subject Contexts (723) 306 157 81 179
Time 2 Age 3;0.4--3;0.18 3;0.1--3;1.27 3;0.0--3;1.15 3;0.0--3;1.11Sessions 3 � 1 h 6 � 1 h 6 � 1 h 6 � 1 hMLU 3.12--3.42 2.5--3.11 2.17--2.49 2.25--2.65N of Subject Context (1028) 191 272 278 287
2.3. Accessibility features
All third person subjects from both the child andmother data that fit the above criteria were coded for accessibility usingsix discourse-pragmatic features: ANIMACY, CONTEXTUAL DISAMBIGUATION, PHYSICAL PRESENCE, PRIOR MENTION, LINGUISTIC
DISAMBIGUATION, and JOINT ATTENTION. These are the same six features that were already assessed individually in Hughes andAllen (2013).
We limited our analysis to subject contexts for several reasons. First, there is a known asymmetry between subjectand object omission in English that has been observed by researchers from different theoretical perspectives (e.g.,Processing: Bloom, 1990; Competence: Hyams and Wexler, 1993; Discourse-pragmatic: Allen, 1997). The issuesinvolved in the realization of subjects vs. objects are substantially different in certain respects, such as the higherlikelihood that objects represent new information. Second, there are very few null objects in our data (i.e., <2%),thereby making it difficult to analyze the full range of forms that we are investigating. Third, restricting the analysis tosubject contexts is in line with much of the previous research on the acquisition of referential forms (e.g., Hyams andWexler, 1993; Orfitelli and Hyams, 2012; Paradis and Navarro, 2003; Salazar Orvig et al., 2010a; Serratrice, 2005;Shin and Cairns, 2012). For all of the reasons stated above, we believe that restricting our analysis to subjectcontexts presents a clearer picture of the effects of discourse-pragmatics than would analyzing subjects and objectstogether.
Each subject in the data was assigned a binary value for each feature based on how accessible its referent was in thediscourse context (e.g., Ariel, 2001; Clancy, 1993; Allen, 2000; Skarabela, 2007b).4 An accessible value indicates that thereferent is easily identified in the discourse context, so the speaker is likely to use a reduced form to realize the referent inspeech (e.g., null form, pronoun). An inaccessible value indicates that the referent is not easily identified in the discoursecontext, so the speaker is likely to use a more informative form to realize the referent in speech (e.g., demonstrative, nounphrase). Table 2 lists the six features and how accessibility vs. inaccessibility was defined for each.
Please cite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
4 In reality, these features are not binary but rather gradient in nature. For example, numerous hierarchies posit degrees of animacy in whichpeople are considered more animate than animals, which are more animate than plants, which in turn are more animate than rocks (Silverstein,1976). A similar gradient can be posited for the other five features as well. However, as it has been the practice in the literature to treat thesefeatures as binary in order to simplify the analysis and to allow for statistical analysis based on these values, the current study maintains the use ofbinary values.
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 7
+ ModelsLINGUA-2160; No. of Pages 19
Table 3Four referential forms -- a hierarchy of accessibility markers.
Referent’s accessibility Form Example
[TD$INLINE]
High
Low
Null Ø want chips. (Eleanor 2;00.15)Pronoun It goes there. (Annie 2;00.25)Demonstrative This go round here. (Fraser 2;00.20)
Lexical NP Boy go. (Brian 2;07.01)
The following example illustrates how each subject was coded. In the interaction in (4), themother and child are playingwith multiple blocks on the table. At the time of the final utterance, the mother is looking at and touching the block inquestion (Block B), and the child is also looking at it.
(4)
Pleason re
5 In thutterancpreviou2008).
MOT:
e cite thferentia
eir corpues prior,s utteran
That block fell down. [referring to Block A]
MOT: Where does this block go? [referring to Block B] CHI: Ø go there. [referring to Block B] (Annie 2;0)
The (omitted) subject in the final utterance is coded as follows:
ANIMACY. A block is inanimate. Because animate entities are particularly salient to children in the discourse context, thesubject that has an inanimate referent is coded as inaccessible.CONTEXTUAL DISAMBIGUATION. At the time of the utterance, the child is playing with several blocks, so there are numerousother possible referents in the physical context that could fit the semantics of the verb and the grammatical elements ofthe intended referent. Therefore, the subject is coded as inaccessible.PHYSICAL PRESENCE. The block is physically present, and therefore more easily identified than if it were absent from thephysical context. The subject is coded as accessible.PRIOR MENTION. Any referent that has been mentioned in the preceding 5 utterances is considered accessible.5 Becausethe block referred to was just mentioned in the previous utterance, it is coded as accessible.LINGUISTIC DISAMBIGUATION. There are two referents mentioned in the preceding five utterances that fit the semantics of theverb and the grammatical elements of the referent, making it more difficult to identify. Therefore, the subject is coded asinaccessible.JOINT ATTENTION. The block is the focus of attention for both the mother and the child, and they are aware of each other’sattention on the block. Thus the subject is coded as accessible.
In order to test inter-rater reliability, 9% of the data was blind-coded by two research assistants. These two codersachieved an average agreement of 85%. Moreover, every file was coded by one research assistant, and thensubsequently checked by a second research assistant. The first author then reviewed all files in order to resolve any inter-rater differences.
2.4. Argument forms
In order to achieve a better understanding of the effects of discourse-pragmatics on subject choice, all third personsubjects in the analysis were categorized as one of four possible referential forms, based on Ariel’s accessibility theory(Ariel, 2001). The central claim of accessibility theory is that adult speakers will choose a referential form based on theaccessibility of the referent. The four referential forms analyzed in this study are null forms, pronouns, demonstratives, andlexical noun phrases. Table 3 demonstrates the hierarchy of these four forms. Lexical forms are low accessibility markersbecause they are in situations when a referent is inaccessible, meaning that the speaker has to convey more informationin order to identify the referent. Conversely, null forms are high accessibility markers because they mark that a referent ishighly accessible in the discourse context, meaning that the speaker does not have to convey as much information to
is article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityl choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
s of Inuktitut child speech, Skarabela and Allen (2003) determined that few arguments have preceding references more than 5showing that the difference between a cut-off of 5 versus 20 preceding arguments is minimal. They suggest that a threshold of 5ces to encode a referent as new rather than given may be an appropriate distance for child spontaneous speech (Allen et al.,
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx8
+ ModelsLINGUA-2160; No. of Pages 19
identify the referent. In other words, the referential form signals the degree of accessibility of a referent. Demonstrativesand pronouns fall in between lexical noun phrases and null forms in terms of accessibility; however, as in many previousstudies, demonstratives are ranked slightly higher than pronouns in terms of informational status, as seen in Table 3(see Ariel, 1990, 2001; Givon, 1983; Gundel et al., 1993 among others).
2.5. Measuring the incremental effect of accessibility
The general claim underpinning this study is that the effect of the accessibility features increases incrementally for bothadults and children, so that a speaker is more likely to produce reduced forms when themajority of features are accessibleand full lexical noun phrases when themajority of features are not accessible. To determine whether this is the case, eachpossible subject context in the data was assigned a binary value -- ‘‘accessible’’ (0) or ‘‘not accessible’’ (1) -- for each of thesix accessibility features. The values of the six features were then summed for each subject, giving it a value between 0(completely accessible in the discourse context) and 6 (completely inaccessible in the discourse context). A two-waycontingency table analysis was then conducted for each participant with the summed value of accessibility as theindependent variable and form (i.e., null, pronominal, demonstrative, lexical) as the dependent variable. Chi-square testsof independence were performed to determine whether there was a statistically significant relationship between thepredictor and the form of the subject.
We maintain the following two hypotheses for our data. First, the cumulative effect of the discourse-pragmatic featureswill mean that both adults and children will be more likely to realize highly accessible referents with reduced linguisticforms and highly inaccessible referents withmore informative forms. Second, children’s choice of linguistic forms will differfrom adults most at Time 1, but will become more adult-like at Time 2, so that their choice of linguistic forms will moreclosely match those of their caregivers.
3. Results
What follows are the results of the analysis for both the children and the adults in this study. In Section 3.1, utterancesfrom the four mothers as a group are analyzed in order to establish how adult caregivers handle the mapping fromaccessibility of a referent to choice of linguistic form in the input and also to determine whether the cumulative effects ofaccessibility can be seen in the adults’ referential choice. In Section 3.2, individual results are given for each of the fourchildren at the two time periods to determine each child’s sensitivity to the incremental effect of accessibility on referentialchoice, and to investigate how this ability develops over time.
Finally, this approach takes as its starting point the assumption that all features have equivalent ‘‘weight’’ indetermining argument realization. This is a necessary assumption in order to perform this kind of analysis; however, inreality, this is not exactly the case. In earlier work, we have investigated the relative weight and the interaction of specificfeatures and found that certain features are stronger than others and affect the choice of referent differently (e.g., Hughesand Allen, 2006; Allen, 2007; Hughes, 2011; Hughes and Allen, 2013). Therefore, in Section 3.3, the relative strength andweight of features in determining referential form are explored for the children as a group at each time point usingmultinomial logistic regression analyses. These findings provide a more refined model of how the six features inform theinterlocutor’s choice of referent.
3.1. Group analysis for the caregiver data
The prediction for the caregiver data is that there should be a strong connection between the accessibility of a referentand referential choice: accessible referents should be realized by reduced linguistic forms (i.e., null forms and pronouns)and inaccessible referents should be realized by informative linguistic forms (i.e., demonstratives and lexical nounphrases). When analyzed individually, the results for the caregivers followed very similar patterns, and no significantchange or development occurred in their production of forms from Time 1 and Time 2. Therefore, the caregivers’ data wascombined as shown in Fig. 1.
The graph in Fig. 1 depicts the cumulative effect of accessibility for six discourse-pragmatic features distributed acrossfour linguistic forms. The accessibility values range from 0 to 6 along the x-axis, with 6 being most inaccessible. Asexpected for the caregiver data, there are very few null subjects overall; instead, the caregivers use pronouns foraccessible referents. The caregivers show the predicted pattern in their data: when referents are completely accessible,they are realized with pronouns at a rate of 67%. The caregivers’ use of pronouns decreases as referents become lessaccessible, and their use of demonstratives and lexical noun phrases increases. When referents are completelyinaccessible for all 6 features, the caregivers produce lexical noun phrases at a rate of 95%.
One thing that was not predicted is the relatively high proportion of lexical noun phrases produced by the caregivers forreferents that were coded as ‘‘inaccessible’’ for only 0 or 1 discourse-pragmatic features. This phenomenon likely results
Please cite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 9
+ ModelsLINGUA-2160; No. of Pages 19
[(Fig._2)TD$FIG]
Fig. 2. Distribution of forms at Time 1 and Time 2 for the children as a group.
[(Fig._1)TD$FIG]
Fig. 1. Cumulative effect of accessibility for six features distributed across four forms for caregivers as a group (all categories significant at thep ≤ .01 level).
from the repetition of lexical noun phrases typical of adult--child discourse that has been documented in several studiessuch as Clark and Bernicot (2008). It will be discussed in more detail in Section 4.
3.2. Individual trends in the child data
Before investigating the hypotheses stated in Section 2.5, the overall distribution of linguistic forms for the four childrenas a group was determined. Fig. 2 represents this distribution for the children at Time 1 (Total N = 723) and Time 2 (TotalN = 1028). Chi-square analyses were performed to compare the use of each form. At Time 2, there are significantly fewernull forms (x2 = 168.470, p < .0001), significantly more pronouns (x2 = 173.103, p < .0001), significantly moredemonstratives (x2 = 7.690, p < .01), and significantly fewer lexical NPs (x2 = 41.060, p < .0001) than at Time 1. Thesefindings are in line with previous studies.
It has beenwell-established that there is considerable variation between individual children in both the rate of languageacquisition and the cognitive strategies that each child applies to the task of language learning (Goldfield and Snow,1997). Therefore, results for the children are reported here individually, in order to exclude the possibility that one or twochildren are responsible for the tendencies seen in the data for the entire group. In Fig. 3a through d, the cumulative effectof accessibility across four linguistic forms is represented for each child at Time 1 and Time 2.
Although there are some differences between the children, they all follow a similar pattern that shows strong sensitivityto the incremental effect of accessibility. At both Time 1 and Time 2, the children generally employ a high proportion ofreduced forms to realize referents that are fully accessible, and a high proportion of more informative forms to realizereferents that are ‘‘not accessible’’ for 4, 5 or 6 features. Moreover, the proportion of reduced forms decreases, and theproportion of more informative forms increases, as the referents become increasingly inaccessible. As expected, thereduced forms used at Time 1 are largely null forms, although some pronouns are used by each child. At Time 2, however,null forms have largely been replaced by pronouns used for the same function. Chi-square tests were conductedcomparing the distribution of forms at Time 1 vs. Time 2 for each child. They show that the distribution of forms differs
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx10
+ ModelsLINGUA-2160; No. of Pages 19
[(Fig._3)TD$FIG]
Fig. 3. (a) Cumulative effect of accessibility for six features distributed across four forms for Annie at Time 1 and Time 2 (all figures significant atthe p ≤ .01 level). (b) Cumulative effect of accessibility for six features distributed across four forms for Brian at Time 1 and Time 2 (all figuressignificant at the p ≤ .01 level). (c) Cumulative effect of accessibility for six features distributed across four forms for Eleanor at Time 1 and Time 2(all figures significant at the p ≤ .01 level). (d) Cumulative effect of accessibility for six features distributed across four forms for Fraser at Time 1and Time 2 (all figures significant at the p ≤ .01 level).
significantly between Time 1 and Time 2when referents are ‘‘inaccessible’’ for 0--4 features (p < .001 for all comparisons).However, when referents are ‘‘inaccessible’’ for 5 or 6 features, the differences were not significant: the children at bothtimes primarily use lexical noun phrases to introduce referents that are fully inaccessible.
As in the adult data, the children produce a higher proportion of lexical noun phrases than expected for referents thatare at the accessible end of the scale, especially at Time 1. This overuse of lexical noun phrases decreases at Time 2 forAnnie, Eleanor, and Fraser, but Brian still relies strongly on lexical noun phrases to realize accessible referents at Time 2.This pattern will be discussed further in Section 4.
The graphs at Time 1 demonstrate that for all the children, themapping from accessibility of the referent to the choice ofargument form is not quite adult-like. For the most part, there is not the neat cline reflecting the accessibility-formrelationship that is seen for the adults in Fig. 1, where reduced linguistic forms dominate the accessible end of the scaleand there is a gradual increase in more informative forms as accessibility increases. However, even at Time 1, it is clearthat the children are sensitive to the discourse context because when referents are inaccessible for 4, 5, and 6 features, allchildren rely heavily on lexical noun phrases. By Time 2, all of the children are producing pronouns, and a similar cline ofaccessibility that we saw in the adult data is seen in the child data, demonstrating that the children have honed theirmapping skills, making obvious their increased sensitivity to the discourse context.
Individual differences are also apparent between the four children. Brian and Fraser use more null forms at Time 1,while Annie and Eleanor use fewer null forms and more demonstratives. This may indicate that the two girls are using
Please cite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 11
+ ModelsLINGUA-2160; No. of Pages 19
more demonstratives instead of null forms and pronouns at Time 1 or that the constructions that they are producing requiredemonstratives; for instance, they may be using more copula constructions for identification than Brian and Fraser. Thechildren’s level of linguistic ability as measured by mean length of utterance also varies slightly, and this may have aneffect. Although the children are age-matched and broadly language-matched, Annie has the highest MLUw at both Time1 (MLUw: 2.45) and Time 2 (MLUw: 3.42) and the pattern of her data at both times looks themost adult-like. Fraser has thelowest MLUw at Time 1 (MLUw: 1.74) and the pattern of his data looks very erratic. Overall, even with these individualdifferences, the graphs in Fig. 3a through d look remarkably alike for the two age ranges.
3.3. The effect of individual features
In the analyses thus far, the six features have been summed and treated as if they have equal weight in determiningreferential form. However, it is more likely the case that each feature does not contribute equally in determining the choiceof referent, In fact, certain features may be stronger than other features in predicting particular outcomes (i.e., null formsvs. pronouns vs. demonstratives vs. lexical NPs).
In order to determine the contribution of each of the six features accurately, multinomial logistic regression analyseswere performed for the children as a group at each of the two time points. Multinomial logistic regression is an advancedstatistical technique that predicts the probability of a particular outcome when there are more than two levels of dependentor outcome variable (e.g., null form, pronoun, demonstrative, or lexical NP) based on a set of independent or predictorvariables (e.g., the six discourse-pragmatic features under investigation). It also can help to determine which of thevariables are strongest in determining the predicted outcome. In the following analyses, all independent variables aredichotomous: a value of ‘0’ was assigned if a referent was accessible for a particular feature, and a value of ‘1’ wasassigned if a referent was inaccessible for that feature. The reference category for all analyses was ‘lexical NP’. In otherwords, the analyses predict whether a speaker is more likely to choose a null subject vs. a lexical NP, a pronoun vs. alexical NP, or a demonstrative vs. a lexical NP based on the values of the predictor variables -- the six discourse-pragmaticfeatures. As stated earlier, the general prediction is that accessible referents are more likely to be represented by lowinformation forms (i.e., null forms and pronouns) and inaccessible referents are more likely to be represented by highinformation forms (i.e., demonstratives and lexical NPs).
3.3.1. Multinomial regression analysis for the children at Time 1A comparison of the model with all six predictors for the child data at Time 1 against a model with no predictors
(intercept-only6) showed significant improvement (df = 18, n = 825, x2 = 313.202, p < .0001). Moreover, likelihood ratiotests showed that all features except for LINGUISTIC DISAMBIGUATION (p = .063) contributed significantly to the model(p < .0001). The model successfully predicted 49.9% of total referential forms overall. For each individual form, itpredicted 76.1% of lexical NPs, 44.2% of null forms, and 36.2% of demonstrative forms correctly. However, it onlypredicted 8.7% of pronominal forms. Overall, the model was successful enough in modeling the children’s choice ofreferential forms at Time 1 that it is worth examining the findings more closely so that we can get a better idea of whatfeatures are important in driving these predictions.
Table 4 shows the effects of the six predictor variables on the choice of referential forms for the children at Time 1. Thepredictor variables for each category are listed in the leftmost column. In each case, the results indicate the likelihood ofthat category occurring as opposed to the reference category, which is lexical NP. The b values in the first column are theestimatedmultinomial logistic regression coefficients for themodel. This is based on the accessible values of the predictorvariables -- a positive b value indicates that a referent is more likely to be realized as a null subject, a pronoun, or ademonstrative than a lexical NP, and a negative b value indicates that a referent is more likely to be realized as a lexicalNP.7 The significance level of the Wald statistic is measured in the column labeled Sig. Anything with a p-value of lessthan .05 is considered significant. Finally, Exp (b) is the odds ratio for the predictors. This number describes how likely acertain outcome is: a value greater than 1 indicates that the outcome will fall into the comparison category (e.g., null ratherthan lexical), while a value less than 1 indicates the outcome will fall into the reference category (e.g., lexical rather thannull). Exp (b) also predicts how many times more likely that outcome will be.
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
6 The intercept is the multinomial logit estimate for each outcome/dependent variable relative to the reference category (lexical NP) when thevalues for the six predictor variables in the model are set to zero; in other words, the intercept-only model shows the outcome when the model isrun without the six discourse-pragmatic features as predictors.
7 The column labeled Std.Error contains the standard errors of the individual regression coefficients for the two models. TheWald Chi-Squarestatistic is the ratio of the coefficient (b) to its standard error squared. The column labeled df lists the degrees of freedom for each of the variables inthe model. In this case, the degree of freedom for each variable is 1.
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx12
+ ModelsLINGUA-2160; No. of Pages 19
Table 4The effect of six predictors (ANIMACY, CONTEXTUAL DISAMBIGUATION, PHYSICAL PRESENCE, LINGUISTIC DISAMBIGUATION, JOINT ATTENTION, and PRIOR MENTION) on thechoice of lexical NP vs. three other possible categories for children at Time 1.
Predictor variables b Std.Error Wald df Sig. Exp (b)
Reference category: lexical NP.* Significant at p < .05.** Significant at p < .01.*** Significant at p < .001.
For the first category, lexical NP vs. null subject, all of the predictor valueswere found to be significant ( p < .05 or lower)except for LINGUISTIC DISAMBIGUATION ( p = .103). The prediction for three of these features, PHYSICAL PRESENCE, JOINT ATTENTION,and PRIOR MENTION, is that if a referent is accessible for any one of these (i.e., physically present, jointly attended to, and/orpreviously mentioned), it is more likely to be realized as a null subject rather than a lexical NP. The strongest predictor isPHYSICAL PRESENCE: according to the odds ratio (Exp (b)) a referent is 4 times as likely to be a null subject if it is accessible forthis feature. A referent that is accessible for PRIOR MENTION or JOINT ATTENTION is more than twice as likely to be realized as anull subject. An accessible value for ANIMACY or CONTEXTUAL DISAMBIGUATION does not have the same predictive value: thesereferents are significantly more likely to be realized as lexical NPs showing that an accessible value for these two featuresdoes not cause the child to produce a null form. Perhaps, instead, the child relies more heavily on English grammaticalconstraints and supplies a subject. Furthermore, at Time 1, the child is also more likely to adhere to the convention of childand caregiver speech described above in which lexical NPs are produced to refer to animate third person entities insteadof null forms or pronouns (see Sections 3.1, 3.2 and 4).
Four of the predictors were found to significantly affect the choice of pronoun vs. lexical NP. An accessible value forPHYSICAL PRESENCE or PRIOR MENTION predicts that a referent is almost three times as likely to be a pronoun rather than alexical NP. An accessible value for ANIMACY or CONTEXTUAL DISAMBIGUATION predicts that a referent is more likely to be realizedas a lexical NP and not a pronoun, again demonstrating that accessible values for these features do not have a strongeffect on the child’s production of low information forms.
Four of the predictors were also found to significantly affect the choice of demonstrative versus lexical NP, but not thesame four. The strongest feature is PHYSICAL PRESENCE: an accessible value for this feature predicts that a referent is morethan 14 times as likely to be realized as a demonstrative and not a lexical NP. A referent that is accessible for JOINT
ATTENTION is over three times as likely to be realized as a demonstrative. These findings are not surprising becausedemonstratives are normally used for objects that are physically present; moreover, they are often a way to draw aninterlocutor’s attention to a specific object. Accessible values for ANIMACY or CONTEXTUAL DISAMBIGUATION predict that areferent is more likely to be realized as a lexical NP and not a demonstrative. As well as indicating that these features arenot as strong, as we have seen earlier, it is also the case that demonstratives are not typically used for animate entities.They are, however, typically used to disambiguate an object in the physical context, so it makes sense that animateentities and unambiguous referents are less likely to be realized by demonstratives and more likely to be realized byanother form, lexical NPs in this case.
Please cite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 13
+ ModelsLINGUA-2160; No. of Pages 19
Table 5The effect of six predictors (ANIMACY, CONTEXTUAL DISAMBIGUATION, PHYSICAL PRESENCE, LINGUISTIC DISAMBIGUATION, JOINT ATTENTION, and PRIOR MENTION) on thechoice of lexical NP vs. three other possible categories for children at Time 2.
Predictor variables b Std.Error Wald df Sig. Exp (b)
Reference category: lexical NP.* Significant at p < .05.** Significant at p < .01.*** Significant at p < .001.
3.3.2. Multinomial regression analysis for the children at Time 2For Time 2, a comparison of the model with all six predictors against a model with no predictors showed significant
improvement (df = 18, n = 1171, x2 = 493.393, p < .0001). Moreover, likelihood ratio tests showed that all featurescontributed significantly to the model (p < .0001). The model also successfully predicted 58.8% of the total referentialforms overall: it predicted 73.0% of pronouns, 62.4% of demonstratives, and 42.0% of lexical NPs correctly. It failed topredict any null forms, but at Time 2, this only represents 4% of the children’s referents. Table 5 describes the effects of thesix predictor variables on the choice of referential forms for the children at Time 2.
None of the predictor values were found to be significant in predicting null subjects except for PRIOR MENTION ( p < .0001).The children at Time 2 produce significantly fewer null subjects than at Time 1, and it seems that the only feature thatinfluences this choice is PRIOR MENTION. In fact, if a referent has been previously mentioned in the last five utterances, it isover five times as likely to be a null subject rather than a lexical NP (Exp (b) = 5.329).
Five out of six predictors were found to significantly affect the choice of pronoun vs. lexical NP. An accessible value forPHYSICAL PRESENCE, LINGUISTIC DISAMBIGUATION, JOINT ATTENTION, or PRIOR MENTION predicts that a referent is more likely to be apronoun and not a lexical NP. JOINT ATTENTION and PRIOR MENTION are the strongest predictors: referents that are accessiblefor either of these features are three times as likely to be realized as a pronoun. However, an accessible value forCONTEXTUAL DISAMBIGUATION predicts instead that a referent is more likely to be realized as a lexical NP and not a pronoun;this again indicates an accessible value for this feature does not play a strong role in determining the production of lowinformation forms.
Four predictors had a significant effect on the choice of demonstrative vs. lexical NP. The strongest feature is PHYSICAL
PRESENCE: an accessible value for this feature predicts that a referent is more than 11 times as likely to be realized as ademonstrative and not a lexical NP. A referent that is accessible for JOINT ATTENTION is almost three times as likely to be ademonstrative. Accessible values for ANIMACY or CONTEXTUAL DISAMBIGUATION predict that a referent is more likely to berealized as a lexical NP and not a demonstrative, most likely for the reasons given in Section 3.3.1.
To summarize, at Time 1, an accessible value for the features PHYSICAL PRESENCE, PRIOR MENTION,or JOINT ATTENTION issignificant in predicting null subjects rather than lexical NPs, while an accessible value for the features PHYSICAL PRESENCE
or PRIOR MENTION is significant in predicting pronouns rather lexical NPs. Lastly, an accessible value for the features PHYSICAL
PRESENCE or JOINT ATTENTION is significant in predicting demonstratives rather than lexical NPs. Therefore, the accessibilityvalue for these three features seems to have the greatest effect on the children’s production of lower information forms anddemonstratives at Time 1.
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx14
+ ModelsLINGUA-2160; No. of Pages 19
By Time 2, the children are more adult-like in their choice of referents, and only one feature is significant in predictingthe occurrence of a null subject rather than a lexical NP --PRIOR MENTION. However, an accessible value for four features --PHYSICAL PRESENCE, LINGUISTIC DISAMBIGUATION, JOINT ATTENTION, and PRIOR MENTION -- predicts that a referent is more likely to be apronoun rather than a lexical NP. Finally, an accessible value for the features PHYSICAL PRESENCE or JOINT ATTENTION issignificant in predicting demonstratives rather than lexical NPs, with PHYSICAL PRESENCE being the strongest predictor. AtTime 2, children’s production of lower information forms is predicted by an accessible value for the same three featuresshown to be significant at Time 1, with the addition of LINGUISTIC DISAMBIGUATION.However, these low information forms areprimarily null forms at Time 1 and pronouns at Time 2.
4. Discussion
In the present study, we examined English-speaking children’s and caregivers’ sensitivity to the incremental effect ofsix discourse-pragmatic features on referential choice at two developmental time periods. Each of the six features hadalready been shown to individually influence referential choice in these same speakers in a previous study (Hughes andAllen, 2013). Here we showed that both the children and their caregivers are sensitive in their referential choice not only tothe individual features but also to the cumulative effect of these features.
Speakers produced a high proportion of reduced forms (i.e., null forms and pronouns) when referents were maximallyaccessible, and a high proportion of informative forms (i.e., lexical noun phrases) when referents were maximallyinaccessible. More importantly, all speakers showed sensitivity to the number of features for which a referent wasinaccessible. The proportion of reduced forms decreased incrementally, and the proportion of informative forms increasedincrementally, as the inaccessibility score of the referent increased. Although there was some variation in the magnitudeof the effect, every participant showed a pattern of this sort. Furthermore, the effect was evident at both developmentalstages that we examined -- Time 1 (2;0--2;7) and Time 2 (3;0--3;1). The children at Time 1 showed this effect most clearlyas referents became increasingly inaccessible. The children at Time 2, like the caregivers, showed a gradient patternacross the scale of accessibility.
The results of this study extend Allen’s (2007) previous findings about incremental sensitivity in Inuktitut to English, alanguage of a very different typology. We further show that there is some development in this sensitivity, with three-year-olds performing more like adults than two-year-olds. Finally, we show that caregivers are sensitive to the incrementaleffect of accessibility as well.
All groups, but especially the children at Time 1, also displayed an interesting anomaly in the gradient pattern: anunexpectedly high proportion of lexical NPs for maximally accessible referents and referents with one inaccessiblefeature. This anomaly is largely due to repetition of a part of the previous utterance, a phenomenon that is quite common inchild--adult interaction. Some examples from our data are in (5).
(5)
Pleason re
8 Repphrases
a.
e citeferent
etitionare us
CHI:
this artial choic
is commoually ref
Butterfly has gone.
icle in press as: Hughes,M.E.,Allen, S.E.M., The incremee in the acquisition of a first language. Lingua (2014), ht
n in adult-to-adult speech as well, and for the same purposes (e.gormulated as pronouns (or null forms) for the second or later men
MOT:
Where has the butterfly gone?[about a puzzle piece] (Fraser, 2;01)
b.
MOT: Is Diesel happy? CHI: Diesel happy. (Fraser, 2;00)
Because these examples represent true third person subjects, they were not removed from the analysis.The high number of noun phrases realizing accessible referents that we found in both child and caregiver data is not
surprising, especially for younger children. In a study of 978 French child--adult exchanges involving repetitions, Clark andBernicot (2008) show that speakers use repetitions for a number of strategic purposes. Adults repeat to check on andconfirm children’s communicative intentions, often in cases where children have not provided complete enoughinformation in their initial utterances to make their meaning clear, as well as to correct errors in what children have said.Both adults and children also use repetitions to signal that they are attending to the other’s utterances, and to place therepeated information in common ground as a basis for further interaction by acknowledging new information from thepreceding speaker’s utterance and showing that they are now treating this as given information (Clark and Bernicot,2008).8 Because children’s speech is often less clear than that of adults, repetition of the full noun phrase rather than just apronoun can be used to clarify what exactly has been said. Repetitions of full noun phrases are also less cognitivelydemanding for children to both comprehend and produce than are pronouns and demonstratives; because lessinformation about the referent is provided in the latter, the interlocutor must infer more information about the referent from
ntal effect of discourse-pragmatic sensitivitytp://dx.doi.org/10.1016/j.lingua.2014.03.001
., Gundel et al., 1993; Walker, 1996). However, nountion of a referent.
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 15
+ ModelsLINGUA-2160; No. of Pages 19
the discourse and physical environment. In addition, repeating the full noun phrase offers children a way to participate inconversation and assume a role in turn-taking before they have fully mastered the more complex system of pronouns anddemonstratives typically used to maintain continuity of reference.
An additional finding related to development involves the use of null vs. pronoun forms for highly accessible referents.Although pronouns are typically used in English to realize accessible referents, the younger children very often producednull forms instead (e.g., ‘‘need hat’’ instead of ‘‘he needs a hat’’). They appropriately select a low information form in thiscontext, but pick an inappropriate or non-adult-like form. This phenomenon of high use of null subjects by English-speaking children has been well-studied for several decades (see review in Hyams, 2011). Again, the complexity of thepronoun system could explain this deviation from the target language. Alternatively, use of null forms could also reflectprocessing difficulty (e.g., Bloom, 1990; Gerken, 1991) or differences in the grammar between children and adults (e.g.,Hyams, 2011; Wexler, 1998).
The present study also shows that pronoun use replaces omitted subjects once the children reach a laterdevelopmental stage. This finding substantiates the assumption in the literature that null forms are used early inacquisition in the place of pronouns (Hyams and Wexler, 1993). The children at Time 1 only produce pronouns forcompletely accessible referents 9% of the time. Instead they use null forms (25% of cases) and lexical noun phrases (64%of cases). The children at Time 2 rarely use null forms to realize accessible referents (only 7% of cases); instead, they nowuse pronouns (60% of cases). This mirrors the reduced forms used by the caregivers to realize accessible referents: 49%pronouns and only 1% null forms. While it is true that the production of null forms is largely replaced by the production ofpronouns, some null formsmight also be taking the place of other forms not fully acquired, such as demonstratives. Thesetwo phenomena in the speech of the younger children -- overuse of lexical NPs and overuse of null forms -- indicating thatthe mapping from accessibility of referent to linguistic form is not completely straightforward for these children.
In a second analysis, we established that the six features did not actually have equal weight in determining referentialform, and by using logistic regression analysis, certain features were demonstrated to be stronger than other features inpredicting particular referential forms.
Results showed that the features PRIOR MENTION, PHYSICAL PRESENCE, and JOINT ATTENTION are strong predictors of aspeaker’s use of low information forms and demonstratives: at Time 1, an accessible value for PHYSICAL PRESENCE is thestrongest predictor of null subjects vs. lexical NPs; accessible values for PRIOR MENTION and PHYSICAL PRESENCE are thestrongest predictors of pronominal subjects vs. lexical NPs; and accessible values for PHYSICAL PRESENCE and JOINT
ATTENTION are the strongest predictors of demonstratives vs. lexical NPs. At Time 2, an accessible value for PRIOR MENTION isthe strongest predictor of null subjects; accessible values for PRIOR MENTION and JOINT ATTENTION are the strongest predictorsof pronominal subjects; and accessible values for PHYSICAL PRESENCE and JOINT ATTENTION are the strongest predictors ofdemonstratives. These findings are in line with our linguistic knowledge of these forms. Demonstratives are normally usedfor objects that are physically present in order to draw the interlocutor’s attention to a specific object without regard to itsstatus as discourse-old or discourse-new, while the use of null subjects or pronouns is more likely to reflect whether areferent has been previously mentioned in the discourse.
As for the three other features, there are varying degrees of significance. LINGUISTIC DISAMBIGUATION does not seem tohave any real predictive power in the model and is rarely significant. However, when it was removed from the analyses,there was no real improvement or change in the models. It may be the case that, although this feature is quite weak whencompared to the other features, it adds predictive strength in combination with other features.
Anaccessible value for ANIMACY is predictive atTime1 for null subjectsvs. lexicalNPsand for pronounsvs. lexicalNPs, butit predictsmembership in the reference category (i.e., lexicalNPs) andnot the comparisoncategory. This ismost likely due tothe fact that only third person pronouns were used in the analysis and the children at Time 1 are using a convention of childand caregiver speech in which they employ lexical NPs instead of null forms or pronouns to refer to animate third personentities. In the category of demonstratives vs. lexical NPs, ANIMACY also predicts membership in the reference category atTime 1 and Time 2 and for the adults. Again, this is as expected because demonstratives are not typically used for animatereferents.
Finally, an accessible value for CONTEXTUAL DISAMBIGUATION does seem to have predictive power for the children at bothtimes; however, like ANIMACY, it predicts membership in the reference category -- lexical NPs. This prediction makes sensein the case of demonstratives as they are often used to disambiguate an object in the physical context; therefore,unambiguous referents are less likely to be realized by demonstratives and more likely to be realized by lexical NPs.However, for the case of the other categories, it merely suggests that this feature does not have strong predictive power forthe production of low information forms.
It is important to note that, while these features do not have equal weight, the model with six predictors showedsignificant improvement over the model with no predictors at both Time 1 and Time 2 and, in most cases, all featurescontributed significantly to the model.
The analyses also show an important developmental change from Time 1 to Time 2. The model at Time 1 was able tocorrectly predict 76.1% of lexical NPs, 44.2% of null forms, and 36.2% of demonstrative forms, but only predicted 8.7% of
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx16
+ ModelsLINGUA-2160; No. of Pages 19
pronominal forms. This supports the findings that children at Time 1 do not yet have the grasp of pronoun usage and areovercompensating by using many lexical NPs. In contrast, the model at Time 2 correctly predicts 73% of pronouns, 62.4%of demonstratives, and 42% of lexical NPs, but does not predict any null forms because the children at Time 2 areproducing very few null forms and instead are producing pronouns in more consistent and adult-like manner. The jumpbetween Time 1 and Time 2 reflects the fact that children at Time 1 are less consistent in their choice of referential formsthan the children at Time 2. As the input becomes more consistent, the model has more predictive power.
While it is true that all six features are not of equal weight and strength, the overall pattern of incremental sensitivitythat we found in this study shows, even more strongly than previous studies based on individual features, that childrenhave some sense of the knowledge of their interlocutors as evident from discourse and contextual features. Although itmight be controversial to label this skill in children as young as 2;0 or 3;0 as a type of ‘mind-reading’ of their interlocutors’intent, there is precedent in recent studies (e.g., Gundel et al., 2007; Gundel and Johnson, 2013; Matthews et al., 2006;O’Neil, 1996, 2005; Wittek and Tomasello, 2005). As Gundel et al. (2007;Gundel and Johnson, 2013) point out, childrenare able to use the full range of referential expressions fairly accurately by the age of 3;0. This ability reflects the fact thatchildren at this early stage are able to gauge the attentional states of their interlocutors. Although the original Theory ofMind (ToM) approach is more concerned with the ability of children to understand the belief systems of others, therudimentary elements of ToM are apparent in the ability of very young children to appropriately use the referentialsystem of the language that they are acquiring. While children may not be able to express the beliefs of theirinterlocutors at earlier ages, they are still able to gauge others’ attention by monitoring expressions, eye gaze, andgesture (e.g., Clements and Perner, 1994; Repacholi and Gopnik, 1997; Robinson andWhitcombe, 2003). In a series ofexperiments, O’Neil (1996, 2005) asserts that very young children’s choice of referent appeals to certain aspects ofToM. In one experiment, she demonstrates that typically developing children as young as two were able to grasp thegiven-new distinction based on the co-presence or absence of their interlocutors when requesting an object that had firstappeared in an earlier event. Children usedmore informative referential forms andmore explicit gestures to indicate theobject when they were aware that their interlocutors were not co-present in some way, either by being out of the room orby shutting both of their eyes at the time of the first presentation of the object (O’Neil, 2005). The children in this study,some as young as 22 months, were able to judge the attentional states of others, which in turn gave them anunderstanding of what constitutes new vs. given in the discourse context, ‘‘leading to a general sense of ‘what peopletalk about’’’ (p. 92).
The present study supports and adds to the growing consensus that very young children have rudimentary aspectsof ToM in place. Gundel et al. (2007) maintain that three-year-old children’s adult-like use of referring expressionsindicates that they have ‘‘the ability to appropriately assess what cognitive status the intended interpretation has for theaddressee at a given point in the discourse’’ (p. 18). In other words, children can determine their interlocutor’sknowledge of the linguistic expression in context. This skill appears at a younger age than previous accounts suggest(see Baron-Cohen, 1995 and Tomasello and Haberl, 2003). However, children at 2;0 and even 3;0 might still might stillnot be able to take the interlocutor’s point of view, and therefore, they might not provide the appropriate information thatis sufficient or necessary for their interlocutor to identify a referent in context (Grice, 1975). The ability to take another’spoint of view seems to develop at a slightly later stage, after 4;0 or even later for certain types of complex deicticreferential expressions and contexts. This can give rise to mapping problems, which are evident in the early use ofreferential forms.
Importantly, the acquisition of discourse pragmatics, in particular the acquisition of referential forms, involves not onlycognitive skills, but also linguistic knowledge. By studying the acquisition and development of referential forms, theinterface between a child’s developing linguistic system and his or her emerging cognitive system becomes evident. Inthis study, we see development along both of these lines from Time 1 to Time 2, marking a distinctive change in thechildren’s production of referential forms by approximately age 3. In terms of linguistic ability, as the children’s MLUsincrease so does their production of more adult-like forms, possibly via constructional learning mechanisms (e.g.,Tomasello, 2003; Lieven et al., 2003) and/or the maturation of innate language mechanisms (e.g., Hyams and Wexler,1993;Wexler, 1998). However, as pointed out above, the children’s correct use of referents is just as dependent on theirdeveloping cognitive abilities -- only this interface between linguistic and cognitive knowledge allows the children tocorrectly gauge the attentional states of their interlocutors and then choose the appropriate linguistic form tocommunicate their intentions.
Overall, the findings here constitute and confirm strong evidence against previous claims that children have adeficient pragmatic system (e.g., De Cat, 2013; Deutsch and Pechmann, 1982; Hamann and Plunkett, 1998; HyamsandWexler, 1993; Schaeffer, 2000; van Hout et al., 2010). Instead, children appear to possess a well-grounded thoughstill developing pragmatic system -- a system that three-year-olds continue to fine-tune on their road to adulthood andthat involves not only linguistic knowledge, but also the coordination of logical, psychological, and social knowledge.This knowledge is not simply a binary differentiation between ‘‘accessible’’ and ‘‘not accessible’’ as we might find in athreshold model. Rather, looking at the incremental effects of accessibility as we have done here reveals a quite
Please cite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 17
+ ModelsLINGUA-2160; No. of Pages 19
subtle sensitivity that enables children to detect small differences in referent accessibility and integrate that withtheir language production. This sensitivity is already evident in children aged 2;0--2;1, and continues to increasewith age.
Acknowledgements
We thank Elena Lieven and Jeannine Goh of the Max Planck Child Study Center in Manchester, UK, for providingaccess to the transcripts and videotapes of the data. This study was funded by National Science Foundation grant BCS-0346841 to Shanley Allen.
References
Allen, S.E.M., 1997. A discourse-pragmatic explanation for the subject-object asymmetry in early null arguments: the Principle of Informativenessrevisited. In: Proceedings of the GALA’97 Conference on Language Acquisition, pp. 10--15.
Allen, S.E.M., 2000. A discourse-pragmatic explanation for argument representation in child Inuktitut. Linguistics 38 (3), 483--521.Allen, S.E.M., 2007. Interacting pragmatic influences on children’s argument realization. In: Bowerman, M., Brown, P. (Eds.), Crosslinguistic
Perspectives on Argument Structure: Implications for Learnability. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 191--210.Allen, S.E.M., Schröder, H., 2003. Preferred argument structure in early Inuktitut spontaneous speech data. In: Du Bois, J.W., Kumpf, L.E., Ashby,
W.J. (Eds.), Preferred Argument Structure: Grammar as Architecture for Function. Benjamins, Amsterdam, pp. 301--338.Allen, S.E.M., Skarabela, B., Hughes, M.E., 2008. Using corpora to examine discourse effects in syntax. In: Behrens, H. (Ed.), Corpora in
Language Acquisition Research: History, Methods, Perspectives. Benjamins, Amsterdam, pp. 99--137.Ariel, M., 1990. Accessing Noun-Phrase Antecedents. Routledge, London.Ariel, M., 1994. Interpreting anaphoric expressions: a cognitive versus a pragmatic approach. Journal of Linguistics 30 (1), 3--42.Ariel, M., 2001. Accessibility theory: an overview. In: Sanders, T.J.M., Schilperoord, J., Spooren, W. (Eds.), Text Representation. Amsterdam,
Benjamins, pp. 29--87.Baron-Cohen, S., 1995. Mindblindedness. An Essay on Autism and Theory of Mind. MIT Press, Cambridge, MA.Bloom, P., 1990. Subjectless sentences in child language. Linguistic Inquiry 21 (4), 491--504.Bock, J.K., Warren, R.K., 1985. Conceptual accessibility and syntactic structure in sentence formulation. Cognition 21, 47--67.Campbell, A., Brooks, P., Tomasello, M., 2000. Factors affecting young children’s use of pronouns as referring expressions. Journal of Speech,
Language, and Hearing Research 43 (6), 1337--1349.Chafe, W., 1987. Cognitive constraints on information flow. In: Tomlin, R. (Ed.), Coherence and Grounding in Discourse. Amsterdam, Benjamins,
pp. 21--51.Clancy, P.M., 1993. Preferred argument structure in Korean acquisition. In: Clark, E.V. (Ed.), Proceedings of the 25th Annual Child Language
Research Forum. CSLI Publications, Stanford, CA, pp. 307--314.Clancy, P.M., 1997. Discourse motivations for referential choice in Korean acquisition. In: Sohn, H., Haig, J. (Eds.), Japanese/Korean Linguistics,
vol. 6. CSLI Publications, Stanford, CA, pp. 639--659.Clark, H.H., 1996. Using Language. Cambridge University Press, Cambridge, MA.Clark, E.V., 2009. First Language Acquisition, 2nd ed. Cambridge University Press, Cambridge.Clark, E.V., Amaral, P.M., 2010.Children build on pragmatic information in language acquisition. LanguageandLinguisticsCompass 4 (7), 445--457.Clark, E.V., Bernicot, J., 2008. Repetition as ratification: how parents and children place information in common ground. Journal of Child Language
35, 349--371.Clements, W.A., Perner, J., 1994. Implicit understanding of belief. Cognitive Development 9, 377--395.De Cat, C., 2004. A fresh look at how young children encode new referents. International Review of Applied Linguistics 42, 111--127.De Cat, C., 2013. Egocentric definiteness errors and perspective evaluation in preschool children. Journal of Pragmatics 56, 58--69.Demir,Ӧ.E., So, W.-C.,Ӧzyürek, A., Goldin-Meadow, S., 2012. Turkish- and English-speaking children display sensitivity to perceptual context in
the referring expressions they produce in speech and gesture. Language and Cognitive Processes 27 (6), 844--867.Deutsch, W., Pechmann, T., 1982. Social interaction and the development of definite descriptions. Cognition 11, 159--184.Du Bois, J.W., 1987. The discourse basis of ergativity. Language 63, 805--855.Gerken, L., 1991. The metrical basis for children’s subjectless sentences. Journal of Memory and Language 30, 431--451.Givon, T. (Ed.), 1983. Topic Continuity in Discourse: A Quantitative Cross-Language Study. Amsterdam, Benjamins.Goldfield, B.A., Snow, C.E., 1997. Individual differences: implications for the study of language acquisition. In: Berko Gleason, J. (Ed.), The
Development of Language. Allyn and Bacon, Needham Heights, MA, pp. 317--347.Graf, E., 2010. An Experimental Pragmatics Approach to Children’s Argument Omissions. (Unpublished doctoral dissertation) Max Planck Child
Study Centre, University of Manchester/Max Planck Institute for Evolutionary Anthropology, Manchester, UK/Leipzig, Germany.Grice, H.P., 1975. Logic and conversation. In: Cole, P., Morgan, J.L. (Eds.), Speech Acts. Academic Press, New York, pp. 41--58.Guerriero, S., Oshima-Takane, Y., Kuriyama, Y., 2006. The development of referential choice in English and Japanese: a discourse-pragmatic
perspective. Journal of Child Language 33 (4), 823--857.Gundel, J.K., Johnson, K., 2013. Children’s use of referring expressions in spontaneous discourse: implications for theory of mind development.
Journal of Pragmatics 56, 43--57.Gundel, J.K., Hedberg, N., Zacharski, R., 1993. Cognitive status and the form of referring expressions in discourse. Language 69 (2), 274--307.Gundel, J.K., Ntelitheos, D., Kowalsky, M., 2007. Children’s use of referring expressions: some implications for theory of mind. ZAS Papers in
Linguistics 48, 1--21.Gürcanli,Ӧ., Nakipoglu, M.,Ӧzyürek, A., 2007. Shared information and argument omission in Turkish. In: Caunt-Nulton, H., Kulatilake, S., Woo, I.
H. (Eds.), Proceedings of the 31st Boston University Conference on Language Development. Cascadilla Press, Somerville, MA, pp. 267--273.
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx18
+ ModelsLINGUA-2160; No. of Pages 19
Hamann, C., Plunkett, K., 1998. Subjectless sentences in child Danish. Cognition 69, 35--72.Huang, C.-c., 2011. Referential choice in Mandarin child language: a discourse-pragmatic perspective. Journal of Pragmatics 43, 2057--2080.Hughes, M.E., 2011. An Analysis of Discourse-Pragmatic andGrammatical Constraints on the Acquisition andDevelopment of Referential Choice
in Child English. (Unpublished doctoral dissertation) Boston University, Boston, MA.Hughes, M.E., Allen, S.E.M., 2006. A discourse-pragmatic analysis of subject omission in child English. In: Bamman, D., Magnitskaia, T., Zaller, C.
(Eds.), Proceedings of the 30th Annual Boston University Conference on Language Development. Cascadilla Press, Somerville, MA, pp.293--304.
Hughes, M.E., Allen, S.E.M., 2013. The effect of individual discourse-pragmatic features on referential choice in child English. Journal ofPragmatics 56, 15--30.
Hyams, N., 1986. Language Acquisition and the Theory of Parameters. Reidel, Dordrecht.Hyams, N., 2008. Reflections on motherese. In: Sano, T., Endo, M., Isobe, M., Otaki, K., Sugisaki, K., Suzuki, T. (Eds.), An Enterprise in the
Cognitive Science of Language: A Festschrift for Yukio Otsu. Hituzu Syobo Publishing, Tokyo, pp. 1--12.Hyams, N., 2011. Missing Subjects in Early child Language. In: De Villiers, J., Roeper, T. (Eds.), Handbook of Generative Approaches to
Language Acquisition. Springer, The Netherlands.Hyams, N., Wexler, K., 1993. On the grammatical basis of null subjects in child language. Linguistic Inquiry 24 (3), 421--459.Küntay, A.C.,Ӧzyürek, A., 2006. Learning to use demonstratives in conversation: what do language specific strategies in Turkish reveal? Journal
of Child Language 33, 303--320.Lieven, E., Behrens, H., Speares, J., Tomasello, M., 2003. Early syntactic creativity: a usage-based approach. Journal of Child Language 30 (2),
333--370.Lieven, E., Salomo, D., Tomasello, M., 2009. Two-year-old children’s production of multiword utterances: a usage-based analysis. Cognitive
Linguistics 20 (3), 481--507.Matthews, D., Lieven, E., Theakston, A., Tomasello, M., 2006. The effect of perceptual availability and prior discourse on young children’s use of
referring expressions. Applied Psycholinguistics 27, 403--422.Mishina-Mori, S., 2007. Argument representation in Japanese/English simultaneous bilinguals: is there a crosslinguistic influence? In: Caunt-
Nulton, H., Kulatilake, S., Woo, I.H. (Eds.), Proceedings of the 31st Annual Boston University Conference on Language Development.Cascadilla Press, Somerville, MA, pp. 441--450.
Narasimhan, B., Budwig, N., Murty, L., 2005. Argument realization in Hindi caregiver-child discourse. Journal of Pragmatics 37 (4), 461--495.O’Neil, D.K., 1996. Two-year-old children’s sensitivity to a parent’s knowledge state when making requests. Child Development 67 (2), 659--677.O’Neil, D.K., 2005. Talking about ‘‘new’’ information: the given/new distinction and children’s developing theory of mind. In: Astington, J.W., Baird,
J.A. (Eds.), Why Language Matters for Theory of Mind. Oxford University Press, pp. 84--105.Orfitelli, R., Hyams, N., 2012. Children’s grammar of null subjects: evidence from comprehension. Linguistic Inquiry 43 (4), 563--590.Paradis, J., Navarro, S., 2003. Subject realization and crosslinguistic interference in the bilingual acquisition of Spanish and English: what is the
role of the input? Journal of Child Language 30 (2), 371--393.Prince, E., 1985. Fancy syntax and ‘shared knowledge’. Journal of Pragmatics 9 (1), 65--81.Repacholi, B.M., Gopnik, 1997. Early reasoning about desires: evidence from 14- and 18-month-olds. Developmental Psychology 33 (1), 12--21.Robinson, E.J., Whitcombe, E., 2003. Children’s suggestibility in relation to their understanding about sources of knowledge. Child Development
74, 48--62.Rozendaal, M.I., Baker, A.E., 2010. The acquisition of reference: pragmatic aspects and the influence of language input. Journal of Pragmatics 42,
1866--1879.Salazar Orvig, A., Marcos, H., Morgenstern, A., Hassan, R., Leber-Marin, J., Parès, J., 2010a. Dialogical beginnings of anaphora: the use of third
person pronouns before the age of 3. Journal of Pragmatics 42, 1842--1865.Salazar Orvig, A., Marcos, H., Morgenstern, A., Hassan, R., Leber-Marin, J., Parès, J., 2010b. Dialogical factors in toddlers’ use of clitic pronouns.
First Language 30, 375--402.Schaeffer, J.C., 2000. The Acquisition of Direct Object Scrambling and Clitic Placement: Syntax and Pragmatics. John Benjamins, Amsterdam.Schmitz, K., 2007. L’interface syntaxe-pragmatique: le sujet chez des enfants bilingues franco-allemands et italo-allemands (The syntax-
pragmatics interface: use of subjects by bilingual French-German and Italian-German children). Acquisition et interaction en langue étrangère25, 9--43.
Serratrice, L., 2005. The role of discourse pragmatics in the acquisition of subjects in Italian. Applied Psycholinguistics 26 (3), 437--462.Serratrice, L., Sorace, A., Paoli, S., 2004. Crosslinguistic influence at the syntax-pragmatics interface: subjects and objects in Italian-English
bilingual and monolingual acquisition. Bilingualism: Language and Cognition 7 (3), 183--205.Shin, N.L., Cairns, H.S., 2012. The development of NP selection in school-age children: reference and Spanish subject pronouns. Language
Acquisition 19, 3--38.Silverstein, M., 1976. Hierarchy of features and ergativity. In: Dixon, R.M.W. (Ed.), Grammatical Categories in Australian Languages. Australian
Institute of Aboriginal Studies, Canberra, Australia, pp. 112--171.Skarabela, B., 2007a. The Role of Social Cognition in Early Language: The Case of Joint Attention in Argument Realization in Child Inuktitut.
(Unpublished doctoral dissertation) Boston University.Skarabela, B., 2007b. Signs of early social cognition in children’s syntax: the case of joint attention in argument realization in child Inuktitut. Lingua
117 (11), 1837--1857.Skarabela, B., Allen, S.E.M., 2002. The role of joint attention in argument realization in child Inuktitut. In: Skarabela, B., Fish, S.A., Do, A.H.-J.
(Eds.), Proceedings of the 26th Annual Boston University Conference on Language Development. Cascadilla Press, Somerville, MA, pp.620--630.
Skarabela, B., Allen, S.E.M., 2003. In: Joint attention in argument expression in child Inuktitut. Paper presented at the Georgetown UniversityRoundtable in Linguistics, Washington, DC.
Skarabela, B., Allen, S.E.M., Scott-Phillips, T.C., 2013. Joint attention helps explain why children omit new referents. Journal of Pragmatics 56,5--14.
Please cite this article in press as: Hughes,M.E.,Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
M.E. Hughes, S.E.M. Allen / Lingua xxx (2014) xxx--xxx 19
+ ModelsLINGUA-2160; No. of Pages 19
So, W.C., Demir, Ӧ.E., Goldin-Meadow, S., 2010. When speech is ambiguous, gesture steps in: sensitivity to discourse-pragmatic principles inearly childhood. Applied Psycholinguistics 31, 209--224.
Tomasello, M., 2003. Constructing a Language: A Usage-based Theory of Language Acquisition. Harvard University Press, Cambridge, MA.Tomasello, M., Haberl, K., 2003. Understanding attention: 12- and 18-month-olds knowwhat is new for other persons. Developmental Psychology
39 (5), 906--912.van Hout, A., Harrigan, K., de Villiers, J., 2010. Asymmetries in the acquisition of definite and indefinite NPs. Lingua 120 (8), 1973--1990.Walker, M.A., 1996. Inferring acceptance and rejection in dialog by default rules of inference. Language and Speech 39, 265--304.Wexler, K., 1998. Very early parameter setting and the Unique Checking Constraint: a new explanation of the optional infinitive stage. Lingua 106,
23--79.Wittek, A., Tomasello, M., 2005. Young children’s sensitivity to listener knowledge and perceptual context in choosing referring expressions.
Applied Psycholinguistics 26, 541--558.
Mary E. Hughes completed her Ph.D. in 2011 in Applied Linguistics at Boston University. Her dissertation focused on the interaction of syntacticand discourse-pragmatic effects in the acquisition of reference in an attempt to shed light on the early development of referential choice in dyadicconversation. Currently, she is a Lecturer at Boston University. Her research interests include the integration of competence-based and usage-based linguistic approaches in order to show the range of factors at work in L1 acquisition. She has also done research on the acquisition of IrishGaelic, and is interested in minority language revitalization.
Shanley E.M. Allen received her Ph.D. in Linguistics from McGill University in Montreal, Canada. She has published extensively on theacquisition of argument realization, the acquisition of morphosyntax and narrative by Inuktitut-speaking children, the coordination of speech andgesture in motion events across typologically different languages, and code mixing in bilingual preschool children. She is Co-Editor of the bookseries Trends in Language Acquisition Research published by Benjamins, and is on the Editorial Board of Bilingualism: Language and Cognition,International Journal of Language Sciences, and Journal of Child Language.
Please cite this article in press as: Hughes,M.E., Allen, S.E.M., The incremental effect of discourse-pragmatic sensitivityon referential choice in the acquisition of a first language. Lingua (2014), http://dx.doi.org/10.1016/j.lingua.2014.03.001
